Abstract: A feature extraction approach of image object's invariant to the translation, scaling and rotation based on wavelet moment is studied. By combining the feature extracting algorithm with BP neural network, an image recognition system is established. Be aim at the quality of image processed is increased, this techniques has priority over other algorithm. In this system, by using wavelet moment invariant, not only the local feature of image object is obtained, but also the description ability for the fine features of image construct is improved. The invariant features extracted from image object are fed into BP neural network to perform the image recognition. In the experiment, using this method to recognize non-noise and noise-added images, especially similar images, the correct recognition rate of the method is up to 98%, besides. Test data attained from the algorithm based on general invariant moment is compared and analyzed. The experimental results demonstrate that the recognizing accuracy and anti-noise capability have been much increased.